Pith. sign in

REVIEW 3 major objections 6 minor 43 references

GLUSE: Enhanced Channel-Wise Adaptive Gated Linear Units SE for Onboard Satellite Earth Observation Image Classification

T0 review · 3 major / 6 minor · reviewed 2026-08-16 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read GLUSE, an SE block augmented with GLU-style gating, claims to beat SE and Gated-SE variants on satellite image classification while keeping the model small enough for onboard deployment.

desk verdict A simple and plausible SE+GLU attention variant for compact satellite-image models, but the 'consistently outperforms' claim rides on single-run comparisons and needs multi-seed verification. read the letter →

arxiv 2504.12484 v1 pith:HUHT5TML submitted 2025-04-16 eess.IV

Thanh-Dung Le , Vu Nguyen Ha , Ti Ti Nguyen , Geoffrey Eappen , Prabhu Thiruvasagam , Hong-fu Chou , Duc-Dung Tran , Hung Nguyen-Kha
show 4 more authors
Luis M. Garces-Socarras Jorge L. Gonzalez-Rios Juan Carlos Merlano-Duncan Symeon Chatzinotas
This is my paper · ORCID
classification eess.IV
keywords channel-wiseattentionsqueeze-and-excitationgatedlinearunitsknowledgedistillationlightweightResNetonboardsatelliteinferenceEarthobservationneuromorphiccomputing
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

ResNet-GLUSE is a lightweight convolutional network for classifying satellite Earth-observation images onboard. The paper's claim is that adding a Gated Linear Unit-style branch to a Squeeze-and-Excitation block, so that channel recalibration combines a static SE rescaling with an input-dependent GLU gate, improves accuracy over plain ResNet, ResNet-SE, and ResNet-Gated-SE. With a 131,565-parameter student distilled from two vision-transformer teachers, the method reaches 94.63% accuracy on EuroSAT and 98.09% on PatternNet, within a few points of MobileViT while using roughly one-33rd of its parameters and one-27th of its FLOPs. The same design runs on a neuromorphic processor at about 852 to 877 mW, which is what makes the claim matter: near-ViT accuracy may be achievable well within satellite power budgets.

What carries the argument

The load-bearing object is the GLUSE block itself: a Squeeze-and-Excitation recalibration $\hat{x} = (x \odot s) + (h \odot g)$, where $s$ is the standard SE channel weight from global-average pooling through two fully connected layers, and $h = W_h * x$, $g = \sigma(W_g * x)$ are two parallel 1×1 convolutions forming a Gated Linear Unit. The SE term contributes global, channel-wise, static weights; the GLU term contributes input-dependent, spatially local gating; the sum makes the recalibration adaptive without adding much complexity, $O(HWC + C^2/r + 2C^2)$.

What would settle it

Re-run the four ResNet variants on EuroSAT and PatternNet across, say, ten seeds with identical splits and hyperparameters; if GLUSE's accuracy lead over Gated-SE (e.g., 0.07 points on PatternNet under KD) is within one standard deviation of the run-to-run differences, the central improvement claim would be unsupported.

Watch

Extended reading notes

Core claim

On its own terms, the paper establishes GLUSE as an architectural upgrade to SE that preserves SE's cheap global recalibration while adding adaptive gating. For input x, GLUSE forms xse = x⊙s with the usual SE weights, computes a GLU pair h = W_h*x and g = σ(W_g*x), forms xglu = h⊙g, and outputs xhat = xse + xglu. The authors report that this sum outperforms both static SE and a Gated-SE variant (which gates only the SE weight s through a 1×1 convolution) across EuroSAT and PatternNet, with and without knowledge distillation. With dual-teacher distillation from EfficientViT and MobileViT, ResNet-GLUSE reaches 94.63% and 98.09% accuracy, respectively, while the MobileViT teacher reaches 99.09% and 99.66%; the claim is that the gap is acceptable given the 33× parameter, 27× FLOP, roughly 6× power, and roughly 3× inference-time savings. A separate deployment on neuromorphic hardware reports 852.30 mW average inference power.

Load-bearing premise

The ranking of methods rests on single-run accuracy differences, so the claim that GLUSE consistently outperforms SE and Gated-SE collapses if those margins are noise rather than systematic effect.

Editorial extensions

If this is right

  • With dual-teacher KD, ResNet-GLUSE reaches 94.63% on EuroSAT and 98.09% on PatternNet, leaving a 4.46- and 1.57-point gap to MobileViT while cutting parameters by 33× and FLOPs by 27×.
  • Without KD, GLUSE still beats the SE baseline (91.05% vs 90.54% on EuroSAT; 88.16% vs 86.34% on PatternNet), so the gating mechanism itself, not just distillation, carries part of the gain.
  • On GPU, the method uses 13.80 W against MobileViT's 79.23 W, and on a neuromorphic processor it runs at about 852–877 mW, 182.42 mJ/frame, and 4.81 fps.
  • The confusion-matrix comparison shows roughly an 8-percentage-point increase in overall accuracy over the plain ResNet baseline on EuroSAT, equivalent to more than 660 additional correct predictions on the 9,000-image test set, concentrated in confused classes such as Highway and River.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Because the GLU branch is input-dependent, GLUSE should be more robust to distribution shift or variable image conditions than static SE; the paper does not test this, but it is a direct consequence of replacing fixed weights with gated, per-sample masks.
  • The dynamic dual-teacher confidence weighting is a separate contribution from GLUSE; ablating it with a fixed 50/50 distillation would isolate how much of the KD result comes from the gating architecture versus the adaptive teacher weighting.
  • The same two-convolution add-on could be dropped into any CNN and, as the paper notes, into ViT backbones; a cheap test is whether a larger ResNet or MobileNet variant gets a similar relative gain with negligible FLOP overhead.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 6 minor

Summary. The paper proposes GLUSE, a modification of the Squeeze-and-Excitation (SE) block that adds GLU-style dual-path gating (a linear 1x1 convolution and a sigmoid-gated 1x1 convolution) on top of the standard SE recalibration, with the two branches combined by addition. The block is inserted into a small ResNet (ResNet8) and evaluated on EuroSAT and PatternNet for Earth observation image classification, both with standard training and with dual-teacher knowledge distillation from EfficientViT and MobileViT. The authors report that ResNet-GLUSE consistently outperforms plain ResNet, ResNet-SE, and a Gated-SE baseline, reaching 94.63% accuracy on EuroSAT and 98.09% on PatternNet under KD, while using 131,565 parameters, 66.56M FLOPs, and 13.80 W GPU power, versus 4.39M parameters, 1.84G FLOPs, and 79.23 W for MobileViT. They further report deployment on the Akida neuromorphic platform with ultralow power consumption (852.30 mW in the abstract/conclusion, 877 mW in Section V).

Significance. If the reported improvements are real, GLUSE would be a cheap, plug-in architectural upgrade that narrows the accuracy gap between lightweight CNNs and ViT teachers at a fraction of the resource cost, with a plausible path to onboard and neuromorphic deployment. The paper's strengths include a clearly specified architecture, evaluation on public benchmark datasets, comparison against independent state-of-the-art baselines (MobileViT, EfficientViT), a public code repository, and detailed parameter/FLOPs/power tables. The central weakness is that all accuracy comparisons rest on single training runs with no variance estimates, so the claimed consistent superiority over SE is not yet statistically supported; the internal inconsistencies in the KD loss definition, the threshold parameter, and the Akida power figure further undermine reproducibility. The potential significance is moderate: the idea is incremental over SE and GLU, but the onboard-satellite and neuromorphic framing gives it practical relevance if the efficiency and robustness claims can be substantiated.

major comments (3)
  1. [Section IV-C and Tables V-VI] The headline claim that ResNet-GLUSE 'consistently outperforms' ResNet-SE and ResNet-Gated-SE is supported only by single-run accuracy point estimates: the experimental protocol in Section IV-C provides no number of seeds, no standard deviations, and no significance tests, and Tables V and VI report exactly one accuracy per model. The most fragile case is PatternNet with KD, where the GLUSE margin over ResNet-SE is 0.07 percentage points (98.09% vs. 98.02%); on a 30,400-image dataset with a 70/30 split, this is well within ordinary seed-level variation. Even the larger EuroSAT KD margin of 1.14 pp needs replication. To make the ranking claim load-bearing, the authors should report mean and standard deviation over at least 3-5 seeds, and ideally a paired significance test or confidence intervals.
  2. [Algorithm 1 vs. Section III-D and Table IV] There are two internal inconsistencies in the dual-teacher KD formulation that directly affect the reproducibility of the KD results in Tables V and VI. First, Algorithm 1 line 8 uses a threshold of 0.4 ('if CT1, CT2 < 0.4 then'), while Table IV lists the confidence threshold δ as 0.6; the subsequent branches also reference δ, so the pseudocode is self-inconsistent. Second, Eq. (18)-(19) define KDloss as a weighted sum of KL divergences where each DKL already includes a 1/τ² factor, but Algorithm 1 line 21 computes KD loss as τ² times the weighted sum of DKL terms, effectively removing the temperature scaling (or double-counting it relative to Eq. 19). The authors should align the pseudocode with the equations and the experiment parameters, and clarify which version was used to produce the reported accuracy numbers.
  3. [Abstract, Section V, and Conclusion] The Akida neuromorphic power consumption is reported inconsistently: the abstract, introduction, and conclusion state 852.30 mW, while Section V reports 'an extremely low inference power consumption averaging 877 mW'. Since ultra-low-power neuromorphic deployment is one of the paper's three stated contributions, this discrepancy is material. The authors should state which measurement is reported in each location, specify the measurement procedure (e.g., whether 852.30 mW is a per-run value and 877 mW is an average across runs), and make the numbers consistent or explain the difference.
minor comments (6)
  1. [Table II and Section III-C] The complexity analysis in Table II understates the cost of the GLUSE convolutions: the 'Linear Conv' and 'Gate Conv' are 1x1 convolutions whose FLOPs scale as HWC², not as C², so their contribution to the total complexity should appear as O(HWC²), not O(C²). The same issue applies to the Gated SE row. Please correct the asymptotic expressions or clarify that the table lists parameter counts rather than FLOPs.
  2. [Equations (22)-(24)] The evaluation metrics are described as 'macro-averaged' in Section IV-C, but Eqs. (22)-(24) define accuracy, precision, and recall with class-size weighting (Nk/N), which is a weighted/micro-style average. The text should be aligned with the formulas, or the formulas should be changed to true macro averaging.
  3. [Section V] The sentence 'achieving metrics closely comparable to the top-performing, with only a slight dip in accuracy' is contradicted by Tables V and VI: on EuroSAT the gap to MobileViT is 4.46 percentage points (94.63% vs. 99.09%), which is not a slight dip. Please rephrase to reflect the actual margins.
  4. [Abstract] The abstract states '≈3x faster inference time (s)', but Table VII gives 6.01 s for ResNet8-GLUSE versus 16 s for MobileViT, which is a 2.66x reduction, not 3x. Please correct the ratio or use '≈2.7x'.
  5. [Fig. 5 and Section V] The text says the boxplot in Fig. 5 'demonstrat[es] stable and high performance across multiple runs', but the caption indicates the boxplot is the inference-side accuracy distribution on Akida, not retraining variance across seeds. Please clarify what the boxplot shows and avoid wording that implies training-seed variability.
  6. [Abstract] There is a grammatical error in the abstract: 'achieving exceeding \textbf{94\% and 98\% accuracy}' should be 'achieving over 94% and 98% accuracy'.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation found: GLUSE's accuracy and efficiency claims are empirical comparisons against external baselines, not consequences of its defining equations.

full rationale

The paper's central claim—that ResNet-GLUSE 'consistently outperforms the standard ResNet, ResNet-SE, and ResNet Gated SE architectures'—is supported by point estimates in Tables V and VI obtained on public datasets (EuroSAT, PatternNet) and compared with independent pretrained models (EfficientViT, MobileViT) as well as same-backbone SE variants. The GLUSE block is defined by Eqs. (6)–(10) as x_se = x⊙s and x_glu = h⊙g, with x_hat = x_se + x_glu; no target accuracy, loss value, or ranking is encoded in these definitions, so the reported 94.63% and 98.09% results are not forced by construction. The self-citations to the authors' prior KD work [14], to [8]/[36] for teacher selection, and to [28] for GLU properties are contextual: the GLUSE-versus-SE comparison is run under identical training settings and does not depend on those citations for its validity. The main weaknesses are non-circular: 'consistently outperforms' rests on single runs with no seeds, confidence intervals, or significance tests (Section IV-C, Tables V–VI), and there are internal inconsistencies (Algorithm 1 threshold 0.4 vs. Table IV δ=0.6; Eqs. (18)–(19) vs. Algorithm 1 line 21; Akida power 877 mW in Section V vs. 852.30 mW in the abstract/conclusion). These are reproducibility and statistical-support concerns, not reductions of the derivation to its inputs.

Assumptions & free parameters 4 free parameters · 3 assumptions · 0 invented entities

The central claims rest on a handful of hand-picked hyperparameters and on the assumption that single-run, single-split numbers are stable. No new physical entities are introduced; the GLUSE block is standard tensor arithmetic. The dynamic KD weighting is new and should have been validated with ablations, especially because Algorithm 1 is internally inconsistent.

free parameters (4)
  • SE reduction ratio r = not reported
    Controls the channel bottleneck in the excitation MLP and appears in the complexity expressions; chosen by hand and not ablated, so the reported parameter and FLOP counts depend on it.
  • KD temperature tau = 5
    Softens teacher and student distributions in Eqs. 14 to 16 and 19; no sensitivity analysis is given, yet it shapes all KD accuracy numbers.
  • KD confidence threshold delta = 0.6 (Table IV; Algorithm 1 line 8 uses 0.4)
    Gates how much each teacher contributes in Algorithm 1; the inconsistency between 0.4 and 0.6 means the implemented weighting may differ from the described one.
  • KD minimum teacher weight w_min = 0.1
    Floor on teacher weights in the dynamic weighting rule; chosen by hand with no ablation.
assumptions (3)
  • domain assumption A single 70/30 split and one training run per configuration yield stable accuracy rankings.
    Tables V and VI and Figures 3 to 4 present one run without seed variance; the claimed consistent superiority of GLUSE depends on this assumption. See Section V.
  • domain assumption GPU power and inference-time measurements are fair and comparable across all models.
    Table VII reports power in watts but no measurement methodology; the headline 6x power claim assumes identical workloads and collection conditions on the RTX 6000.
  • ad hoc to paper The dual-teacher dynamic weighting objective in Algorithm 1 is correctly implemented and equivalent to Eqs. 14 to 21.
    Algorithm 1 line 8 uses a different threshold than the text, and the tau-squared placement differs between Eqs. 18 to 19 and Algorithm 1; if the implementation follows one form, the formal description is wrong.

how reviews work

0 comments
Cite this review

Pith. "Pith review of GLUSE: Enhanced Channel-Wise Adaptive Gated Linear Units SE for Onboard Satellite Earth Observation Image Classification." pith.science (2026). https://pith.science/paper/HUHT5TML

@misc{pith2026250412484,
  author       = {Pith},
  title        = {Pith review of: GLUSE: Enhanced Channel-Wise Adaptive Gated Linear Units SE for Onboard Satellite Earth Observation Image Classification},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HUHT5TML}},
  note         = {Machine review of arXiv:2504.12484}
}
abstract

This study introduces ResNet-GLUSE, a lightweight ResNet variant enhanced with Gated Linear Unit-enhanced Squeeze-and-Excitation (GLUSE), an adaptive channel-wise attention mechanism. By integrating dynamic gating into the traditional SE framework, GLUSE improves feature recalibration while maintaining computational efficiency. Experiments on EuroSAT and PatternNet datasets confirm its effectiveness, achieving exceeding \textbf{94\% and 98\% accuracy}, respectively. While \textbf{MobileViT achieves 99\% accuracy}, ResNet-GLUSE offers \textbf{33x fewer parameters, 27x fewer FLOPs, 33x smaller model size (MB), $\approx$6x lower power consumption (W), and $\approx$3x faster inference time (s)}, making it significantly more efficient for onboard satellite deployment. Furthermore, due to its simplicity, ResNet-GLUSE can be easily mimicked for \textbf{neuromorphic computing}, enabling ultra-low power inference at just \textbf{852.30 mW} on Akida Brainchip. This balance between high accuracy and ultra-low resource consumption establishes ResNet-GLUSE as a practical solution for real-time Earth Observation (EO) tasks. Reproducible codes are available in our shared repository.

Figures

Figures reproduced from arXiv: 2504.12484 by the authors.

Figure 1
Figure 1. A lightweight ResNet backbone, ResNet-GLUSE Module, and GLUSE snipet code in Pytorch. [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. The schematic workflow of dynamic weighting in dual-teacher KD. [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Training (Top) and validation (Bottom) learning curve from standard training strategy. TABLE VII: Model Comparison on Parameters, FLOPs, Size, Inference Time, and Power Consumption [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Training (Top) and validation (Bottom) learning curve with KD training strategy. [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 5
Figure 5. Figure 5: Performance of the ResNet-GLUSE during the inference on [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]
Figure 6
Figure 6. Figure 6: Qualitative results of different channel-wise attention for the EuroSat, and PatternNet data by using Grad-CAM. [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]
Figure 7
Figure 7. Figure 7: Confusion matrix from ResNet (left) and ResNet-GLUSE with KD (right) on the EuroSat dataset. [PITH_FULL_IMAGE:figures/full_fig_p011_7.png]
Figure 8
Figure 8. Figure 8: Confusion matrix from ResNet-GLUSE with KD on the PatternNet dataset. [PITH_FULL_IMAGE:figures/full_fig_p012_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

43 extracted references · 33 canonical work pages

  1. [1]

    New satellite market forecast anticipates 1,700 satellites to be launched on average per year by 2030 as new entrants and incumbents increase their investment in space,

    S. Sadek, “New satellite market forecast anticipates 1,700 satellites to be launched on average per year by 2030 as new entrants and incumbents increase their investment in space,” [Online]. Available: https://shorturl. at/6PVma, 2021

    work page 2021

  2. [2]

    A survey on nongeostationary satellite systems: The communication perspective,

    H. Al-Hraishawi, H. Chougrani, S. Kisseleff, E. Lagunas, and S. Chatzinotas, “A survey on nongeostationary satellite systems: The communication perspective,” IEEE Commun. Surveys & Tuts. , vol. 25, no. 1, pp. 101–132, 2022

    work page 2022

  3. [3]

    Connecting space missions through ngso con- stellations: feasibility study,

    H. Chougrani, et al. , “Connecting space missions through ngso con- stellations: feasibility study,” Frontiers in Commun. Netw. , vol. 5, p. 1356484, 2024

    work page 2024

  4. [4]

    Artificial Intelligence for Satellite Communication and Non-Terrestrial Networks: A Survey

    G. Fontanesi, et al. , “Artificial intelligence for satellite commu- nication and non-terrestrial networks: A survey,” arXiv preprint arXiv:2304.13008, 2023

    work page Pith review arXiv 2023

  5. [5]

    The ϕ-sat-1 mission: The first on-board deep neural network demonstrator for satellite earth observation,

    G. Giuffrida, et al. , “The ϕ-sat-1 mission: The first on-board deep neural network demonstrator for satellite earth observation,”IEEE Trans. Geosci. Remote Sens. , vol. 60, pp. 1–14, 2021

    work page 2021

  6. [6]

    Artificial intelligence based on-board image compression for the ϕ-sat-2 mission,

    G. Guerrisi, F. Del Frate, and G. Schiavon, “Artificial intelligence based on-board image compression for the ϕ-sat-2 mission,” IEEE J. Sel. Top. Appl. Earth. Obs. Remote. Sens. , 2023

    work page 2023

  7. [7]

    Semantic inference-based deep learning and mod- eling for earth observation: Cognitive semantic augmentation satellite networks,

    H. F. Chou, et al. , “Semantic inference-based deep learning and mod- eling for earth observation: Cognitive semantic augmentation satellite networks,” arXiv preprint arXiv:2409.15246 , 2024. IEEE, VOL., NO., 2025. 11 Fig. 7: Confusion matrix from ResNet (left) and ResNet-GLUSE with KD (right) on the EuroSat dataset

    arXiv 2024

  8. [8]

    On-board satellite image classification for earth observation: A comparative study of vit models,

    T. D. Le, et al. , “On-board satellite image classification for earth observation: A comparative study of vit models,” arXiv preprint arXiv:2409.03901, 2024

    arXiv 2024

Show all 43 references
  1. [9]

    Balancing practical uses and ethical concerns: The role of large language models in scientific research,

    R. Ramachandran and K. Bugbee, “Balancing practical uses and ethical concerns: The role of large language models in scientific research,” Per- spectives of Earth and Space Scientists, vol. 6, no. 1, p. e2024CN000258, 2025

    2025

  2. [10]

    A comparison between vgg16, vgg19 and resnet50 architecture frameworks for image classification,

    S. Mascarenhas and M. Agarwal, “A comparison between vgg16, vgg19 and resnet50 architecture frameworks for image classification,” in 2021 International Conference on Disruptive Technologies for Multi- disciplinary Research and Applications , vol. 1, 2021, pp. 96–99

    2021

  3. [11]

    Battle of the backbones: A large-scale comparison of pretrained models across computer vision tasks,

    M. Goldblum, et al., “Battle of the backbones: A large-scale comparison of pretrained models across computer vision tasks,” Adv. Neural Inf. Process. Syst., vol. 36, 2024

    2024

  4. [12]

    Exploring the synergies of hybrid convolutional neural network and vision transformer architectures for computer vision: A survey,

    Y . Haruna, et al., “Exploring the synergies of hybrid convolutional neural network and vision transformer architectures for computer vision: A survey,” Engineering Applications of Artificial Intelligence , vol. 144, p. 110057, 2025

    2025

  5. [13]

    Distilling the knowledge in a neural network,

    G. Hinton, “Distilling the knowledge in a neural network,” arXiv preprint arXiv:1503.02531, 2015

    2015 arXiv

  6. [14]

    Semantic knowledge distillation for onboard satel- lite earth observation image classification,

    T.-D. Le, et al. , “Semantic knowledge distillation for onboard satel- lite earth observation image classification,” 2025 IEEE International Conference on Machine Learning for Communication and Networking (ICMLCN), 2025

    2025

  7. [15]

    GLU variants improve transformer,

    N. Shazeer, “GLU variants improve transformer,” arXiv preprint arXiv:2002.05202, 2020

    2002 arXiv

  8. [16]

    Squeeze-and-excitation networks,

    J. Hu, L. Shen, and G. Sun, “Squeeze-and-excitation networks,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2018, pp. 7132–7141

    2018

  9. [17]

    Eurosat: A novel dataset and deep learning benchmark for land use and land cover classification,

    P. Helber, B. Bischke, A. Dengel, and D. Borth, “Eurosat: A novel dataset and deep learning benchmark for land use and land cover classification,” IEEE J. Sel. Top. Appl. Earth. Obs. Remote Sens., vol. 12, no. 7, pp. 2217–2226, 2019

    2019

  10. [18]

    Patternnet: A benchmark dataset for performance evaluation of remote sensing image retrieval,

    W. Zhou, S. Newsam, C. Li, and Z. Shao, “Patternnet: A benchmark dataset for performance evaluation of remote sensing image retrieval,” ISPRS journal of photogrammetry and remote sensing , vol. 145, pp. 197–209, 2018

    2018

  11. [19]

    Adaptive feature recombination and recalibration for semantic segmentation with fully convolutional networks,

    S. Pereira, et al. , “Adaptive feature recombination and recalibration for semantic segmentation with fully convolutional networks,” IEEE transactions on medical imaging , vol. 38, no. 12, pp. 2914–2925, 2019

    2019

  12. [20]

    Aggregated dense layer in squeeze and excitation networks,

    M. Narayanan, “Aggregated dense layer in squeeze and excitation networks,” in Intelligent Systems Conference. Springer, 2024, pp. 510– 525

    2024

  13. [21]

    Gcnet: Non-local networks meet squeeze-excitation networks and beyond,

    Y . Cao, J. Xu, S. Lin, F. Wei, and H. Hu, “Gcnet: Non-local networks meet squeeze-excitation networks and beyond,” in Proceedings of the IEEE/CVF international conference on computer vision workshops , 2019, pp. 0–0

    2019

  14. [22]

    Training spiking neural networks using lessons from deep learning,

    J. K. Eshraghian, et al., “Training spiking neural networks using lessons from deep learning,” Proceedings of the IEEE, vol. 111, no. 9, pp. 1016– 1054, 2023

    2023

  15. [23]

    The neurobench framework for benchmarking neuro- morphic computing algorithms and systems,

    J. Yik, et al. , “The neurobench framework for benchmarking neuro- morphic computing algorithms and systems,” Nature Communications, vol. 16, no. 1, p. 1545, 2025

    2025

  16. [24]

    Energy-efficient on-board radio resource manage- ment for satellite communications via neuromorphic computing,

    F. Ortiz, et al. , “Energy-efficient on-board radio resource manage- ment for satellite communications via neuromorphic computing,” IEEE Transactions on Machine Learning in Communications and Networking, vol. 2, pp. 169–189, 2024

    2024

  17. [25]

    Language modeling with gated convolutional networks,

    Y . N. Dauphin and et. al., “Language modeling with gated convolutional networks,” in International Conference on ML , 2017, pp. 933–941

    2017

  18. [26]

    Gated transformer networks for multivariate time series classification,

    M. Liu and et. al., “Gated transformer networks for multivariate time series classification,” arXiv preprint arXiv:2103.14438 , 2021

    2021 arXiv

  19. [27]

    Temporal fusion transform- ers for interpretable multi-horizon time series forecasting,

    B. Lim, S. ¨O. Arık, N. Loeff, and T. Pfister, “Temporal fusion transform- ers for interpretable multi-horizon time series forecasting,” International Journal of Forecasting, vol. 37, no. 4, pp. 1748–1764, 2021

    2021

  20. [28]

    Transformer meets gated residual networks to enhance photoplethysmogram artifact detection informed by mutual information neural estimation,

    T.-D. Le, “Transformer meets gated residual networks to enhance photoplethysmogram artifact detection informed by mutual information neural estimation,” arXiv preprint arXiv:2405.16177 , 2024

    2024 arXiv

  21. [29]

    Empirical analysis of squeeze and excitation-based densely connected cnn for chili leaf disease identification,

    N. V ., Y . G., N. N. B., M. R., and P. P., “Empirical analysis of squeeze and excitation-based densely connected cnn for chili leaf disease identification,” IEEE Transactions on Artificial Intelligence, vol. 5, no. 4, pp. 1681–1692, 2024

    2024

  22. [30]

    Lightweight spectral–spatial squeeze-and- excitation residual bag-of-features learning for hyperspectral classification,

    S. K. Roy, S. Chatterjee, S. Bhattacharyya, B. B. Chaudhuri, and J. Platoˇs, “Lightweight spectral–spatial squeeze-and- excitation residual bag-of-features learning for hyperspectral classification,” IEEE Transac- tions on Geoscience and Remote Sensing , vol. 58, no. 8, pp. 52...

    2020

  23. [31]

    Eca-net: Efficient channel attention for deep convolu- tional neural networks,

    Q. Wang, et al., “Eca-net: Efficient channel attention for deep convolu- tional neural networks,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2020, pp. 11 534–11 542

    2020

  24. [32]

    A survey on efficient vision transformers: algorithms, techniques, and performance benchmarking,

    L. Papa, P. Russo, I. Amerini, and L. Zhou, “A survey on efficient vision transformers: algorithms, techniques, and performance benchmarking,” IEEE Trans. Pattern Anal. Mach. Intell. , 2024

    2024

  25. [33]

    Does knowledge distillation really work?

    S. Stanton, P. Izmailov, P. Kirichenko, A. A. Alemi, and A. G. Wilson, “Does knowledge distillation really work?” Adv. Neural Inf. Process. Syst., vol. 34, pp. 6906–6919, 2021

    2021

  26. [34]

    R ´enyi divergence and kullback-leibler divergence,

    T. Van Erven and P. Harremos, “R ´enyi divergence and kullback-leibler divergence,” IEEE Transactions on Information Theory , vol. 60, no. 7, pp. 3797–3820, 2014

    2014

  27. [35]

    Knowledge distillation IEEE, VOL., NO., 2025. 12 Fig. 8: Confusion matrix from ResNet-GLUSE with KD on the PatternNet dataset. from a stronger teacher,

    T. Huang, S. You, F. Wang, C. Qian, and C. Xu, “Knowledge distillation IEEE, VOL., NO., 2025. 12 Fig. 8: Confusion matrix from ResNet-GLUSE with KD on the PatternNet dataset. from a stronger teacher,” Adv. Neural Inf. Process. Syst. , vol. 35, pp. 33 716–33 727, 2022

    2025

  28. [36]

    A semantic-loss function modeling frame- work with task-oriented machine learning perspectives,

    T. T. Nguyen, et al. , “A semantic-loss function modeling frame- work with task-oriented machine learning perspectives,” arXiv preprint arXiv:2503.09903, 2025

    2025 arXiv

  29. [37]

    Efficientvit: Memory efficient vision transformer with cascaded group attention,

    X. Liu, et al. , “Efficientvit: Memory efficient vision transformer with cascaded group attention,” in Proc. IEEE Comput. Soc. Conf. Comput. Vis. Pattern Recognit., 2023, pp. 14 420–14 430

    2023

  30. [38]

    Separable self-attention for mobile vision transformers,

    S. Mehta and M. Rastegari, “Separable self-attention for mobile vision transformers,” Trans. Mach. Learn. Res. , vol. 2023, 2023. [Online]. Available: https://openreview.net/forum?id=tBl4yBEjKi

    2023

  31. [39]

    Scikit-learn: Machine learning in Python,

    F. Pedregosa and et. al, “Scikit-learn: Machine learning in Python,” J. Mach. Learn. Res. , vol. 12, pp. 2825–2830, 2011

    2011

  32. [40]

    Autoaug- ment: Learning augmentation strategies from data,

    E. D. Cubuk, B. Zoph, D. Mane, V . Vasudevan, and Q. V . Le, “Autoaug- ment: Learning augmentation strategies from data,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2019, pp. 113–123

    2019

  33. [41]

    A probabilistic interpretation of precision, recall and f-score, with implication for evaluation,

    C. Goutte and et. al., “A probabilistic interpretation of precision, recall and f-score, with implication for evaluation,” in European Conference on Information Retrieval . Springer, 2005, pp. 345–359

    2005

  34. [42]

    Accessed: March 13, 2025

    Akida Brainchip. Accessed: March 13, 2025. [Online]. Available: https://brainchip.com/akida-neural-processor-soc/

    2025

  35. [43]

    Grad-CAM: Visual explanations from deep networks via gradient-based localization,

    R. R. Selvaraju, et al. , “Grad-CAM: Visual explanations from deep networks via gradient-based localization,” in Proceedings of the IEEE international conference on computer vision , 2017, pp. 618–626

    2017

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.